AI-Based OCT Image Analysis Identifies High-Risk Plaques in Coronary Arteries

While optical coherence tomography (OCT) provides detailed images of vessel structure during coronary interventions, it does not directly reveal the composition of the vessel wall. Identifying dangerous plaques, therefore, often depends heavily on a physician's experience and visual interpretation. Researchers have now developed an artificial intelligence (AI)-based method that detects and maps lipid deposits within OCT images, potentially enabling earlier identification of high-risk plaques.

Researchers at Korea Advanced Institute of Science and Technology (KAIST, Daejeon, South Korea), in collaboration with Korea University Guro Hospital (Seoul, South Korea), have developed a technique to extract wavelength-dependent spectral information embedded within OCT signals and integrate it into a deep learning framework. Because different tissues absorb and reflect light differently, lipid, fibrous tissue, and calcium produce distinct optical signatures. The AI model analyzes these subtle spectral patterns and automatically highlights regions likely to contain lipid-rich plaques, without requiring hardware modifications to existing clinical OCT systems.

Unlike conventional AI systems that require detailed pixel-level annotations, the new approach learns from simpler frame-level labels indicating whether lipid is present. This significantly reduces the burden of data labeling and improves practicality for clinical use. Validation using intravascular imaging data from a rabbit model of atherosclerosis showed strong classification performance and good spatial agreement with histopathology findings. The results, published in Biomedical Optics Express, demonstrate accurate detection of lipid-rich plaques.

The system could provide additional information during coronary interventions to support risk assessment, treatment planning, and evaluation of therapeutic response. By working with existing OCT platforms, the method offers a scalable pathway toward clinical integration. Researchers are now refining processing speed and robustness to enable real-time use and plan further validation using human coronary artery data. They also aim to explore the adaptation of the framework to other intravascular and optical imaging technologies.

“During a coronary intervention, this method could provide clinicians with additional information to support risk assessment, procedural planning, and evaluation of treatment response,” said research team leader Hyeong Soo Nam from KAIST. “Ultimately, it has the potential to contribute to safer clinical decision making, more individualized treatment strategies, and improved long-term management of patients with coronary artery disease.”

Related Links:
KAIST
Korea University Guro Hospital